ARTIFICIAL INTELLIGENCE-BASED HYBRID DEEP LEARNING FRAMEWORK FOR AUTOMATED BONE CANCER DETECTION AND MULTI-STAGE CLASSIFICATION USING RADIOLOGICAL IMAGING

Abstract

Bone cancer is a relatively uncommon but highly aggressive musculoskeletal malignancy that requires timely diagnosis to improve treatment outcomes and patient survival. Radiological imaging modalities, including X-ray, computed tomography (CT), and magnetic resonance imaging (MRI), remain fundamental for diagnosis; however, manual interpretation is time-consuming and may be influenced by inter-observer variability. Recent advances in artificial intelligence (AI) and deep learning have demonstrated significant potential for improving the accuracy and efficiency of computer-aided diagnostic systems. This study presents a hybrid deep learning framework for automated bone cancer detection and multi-stage classification using radiological imaging. The proposed framework incorporates image preprocessing, region-of-interest segmentation, handcrafted texture feature extraction, and hybrid deep learning-based classification to distinguish healthy bone tissue from malignant tumors across four clinical stages (Stage I-IV). A dataset comprising 2,500 radiological images obtained from publicly available bone tumor repositories and collaborating healthcare institutions was used for model development and evaluation. Experimental results demonstrated an overall classification accuracy of 98.41%, sensitivity of 97.92%, specificity of 98.76%, precision of 98.18%, F1-score of 98.05%, and an area under the receiver operating characteristic curve (AUC) of 0.991. Comparative evaluation against conventional machine learning models, including Random Forest, K-Nearest Neighbors, and Support Vector Machine classifiers, showed improved diagnostic performance of the proposed framework. The proposed approach has the potential to support clinicians by facilitating early detection, accurate disease staging, and timely therapeutic decision-making. These findings suggest that AI-assisted diagnostic systems may enhance computeraided assessment of bone malignancies and contribute to improved clinical workflow and patient management in orthopedic oncology.